To support the goals of image guided neurosurgery, we plan to continue our work in developing novel computer vision systems that: segment medical scans into graphical models of distinct anatomical structures; register such models with the position of the patient in the operating room; provide visualization tools that let the surgeon see the full panoply of the patient's anatomy while at the same time supporting surgical access through minimally invasive openings; track surgical instruments relative to the patient and provide the surgeon with a visualization of the position of the instrument with respect to the segmented scans; and automatically adjust models to reflect deformations and changes as the surgery proceeds. In the area of segmentation, we will continue to develop clinically relevant methods for automatically or semi-automatically constructing anatomical models from standard medical scans. The goal is to build patient- specific models that directly highlight structures of interest to the surgeon. In a related sub-proposal, we provide a detailed description of novel methods to be developed to support such segmentation techniques. Within the context of this sub-proposal, we intend to: Extend our current work in segmentation by coupling tissue classification methods with anatomical atlases. This will allow us to use predefined models to serve as templates for guiding the extraction of structures from scans Evaluate the accuracy and stability of existing and modified segmentation techniques. This will include baseline studies that compare our segmentation results against those of expert radiologists, and will include verification of the accuracy of our segmented results against actual anatomy observed during the surgical procedure. Create automatic tools that directly support Image Guided Surgery in neurosurgery, by bringing the segmented models into direct relationship with surgical patients, for use in guidance and navigation. This will include evaluation by the surgeon of what aspect of segmented models are more directly relevant and useful in the surgical procedure.